Sunday, July 6, 2014

Marking your card for tomorrow

There's a new referendum poll out tomorrow in the Sunday Post.  If it follows the same pattern as the last poll in the paper on Easter Sunday, the results won't appear online until relatively late in the morning, in which case I may not be able to post anything about them for a few hours.  So to tide you over, here is the little I know at time of writing -

1) The No camp's embarrassment of a campaign chief Blair McDougall has tweeted about there being a poll tomorrow, so if he's referring to the one in the Sunday Post it must be at least spinnable as good for No - although that could mean anything from a statistically irrelevant margin-of-error increase in the No lead to something more substantial.  (I haven't seen any more immature gloating from the No campaign's "Director of Communications" Rob Shorthouse yet, although that probably just means that he's out drinking again.)

2) The last Sunday Post poll was conducted by Survation.  That doesn't necessarily mean this one is another Survation poll, but they'd be the most obvious suspect.  If so, any change will be from the wafer-thin 5% No lead recorded in the company's last poll - an all-time low.  As I pointed out at the time, the last poll was slightly questionable because 16-24 year olds were breaking unusually well for Yes, and that part of the sample had been very sharply upweighted, meaning that any error caused by random sampling variation would have been magnified.  So when you bear that in mind, a reversion to the mean in this new poll (assuming it is from Survation) would not be at all surprising.  An 8-10 point lead for No has been much more typical for Survation.

3) A Sunday Post journalist has tweeted that the poll shows "some unexpected findings".  One way of interpreting that (although this is just wild speculation) is that the increase in the No lead is over and above a reversion to the mean, which might be said to be surprising in the sense that this week's YouGov v Survation dispute was predicated on the assumption that Survation are a much more Yes-friendly pollster than YouGov.  But even if this does turn out to be a Survation poll with a No lead above the company's normal range of 8-10 points, it's important not to jump to conclusions.  It's perfectly possible that random sampling variation could cause one poll with an unusually low No lead to be followed by one with an unusually high No lead.  That's exactly what happened with ICM recently - a low lead of 3 was followed by a high lead of 12, and the next month it settled back down to a more typical figure of 7.  In reality, all of those apparent shifts in opinion were probably illusions.

UPDATE : An anti-independence troll on Twitter has claimed that it's a TNS poll rather than Survation.  I'm inclined to say that seems implausible, although I wouldn't completely dismiss the idea out of hand, because it's amazing how many Twitter trolls turns out to be salaried, in-the-loop No campaign staffers.

UPDATE II : Mystery solved - there's a TNS poll out, but on Tom Hunter's website.  Survation remain the favourites for the Sunday Post, who I imagine would be unlikely to simply recycle figures from Hunter's website (although you never know, I suppose).

No comments:

Post a Comment